The Power and Pitfalls of Using Multiples to Value Companies


What matters to company owners, of course, is actual cash flows over the long-term future for their business (i.e. the “forever” horizon).  When deciding whether to buy that business today, strict fundamental investors also consider the likely cash flows they may capture over the “forever” horizon.  What determines the price they must pay to invest, however, is expectations for those long-term cash flows as of today, as determined by the marginal investor.  Later, say in five years, if they want to sell that business or raise capital, then expectations for cash flows from year six to “forever” will influence the price they receive.  While the forever horizon is nice, most investors who are seeking an eventual exit, as well as all companies that may require future capital to grow at some point before “forever,” care both about the actual cash flows and about how expectations for those long-term cash flows may affect valuation in the future.

Forever is a long time, and there are many key drivers that could change an awful lot during that time.  There is enormous value in mapping out how future cash flows may unfold, particularly across multiple scenarios, if it can be done in a logical and efficient way.  Even if we can reasonably estimate these cash flows for many years, however, discounted cash flow (“DCF”) has gotten a bad name in many circles because it requires “too many assumptions” and depends a lot on the discount rate one chooses for the DCF.  It also takes a lot of work to produce and communicate, even for a single base case.

As a result, many managers and investors use multiples to determine the long-term value of a business.  It is generally easy to observe any number of fundamental financial measures (like revenue) for the prior year, quite commonplace to predict these measures with confidence for the upcoming year, and often feasible to estimate them a few years out.  Many managers and investors also have confidence in their ability to forecast the “right” multiple to apply to that measure based on a peer group of comparables (“comps”) and the company’s outlook for key things like scale/stability,  growth outlook, and eventual peak profitability. With those two strokes, simple multiplication solves the valuation puzzle:  

Fundamental Financial Measure x Multiple Applied = Valuation

Beyond simplicity, the single biggest reason most investors use multiples is that everyone else uses multiples:  You are generally buying from one who uses multiples and later selling to another who also uses multiples. Investors who have used multiples in their valuation decision-making for some time have a comfortable confidence about what ranges “make sense” based on the characteristics of a business:  They rely on their business knowledge and pattern recognition skills to ensure that the company they are valuing is compared to the “right set of comps,” and they adjust the multiple upwards for favorable characteristics - such as higher growth, better profit margins, longer period of dominance, larger scale/stability/competitive moat - and downwards for unfavorable factors relative to comps.  Organizations also gravitate around multiples: The analyst knows what multiple the partner is comfortable paying, the CEO understands what multiple his board will accept for an acquisition, and everyone “knows” where the comps have traded. It is a lovely, well-ordered way to think about the world.

Let’s take a closer look at each step.

To which fundamental financial measure should the multiple be applied?

Revenue, earnings before interest, taxes & depreciation (“EBITDA”), free cash flow (“FCF”), or earnings per share (“P/E”)?  It depends on the stage and nature of the business. “Everyone knows” that revenue multiples are best for high growth business without earnings, EBITDA fits well for a consolidating roll-up of cash flowing companies, and P/E is the tool of choice for most public stocks that are expected to produce profits for many years to come.  Sometimes, of course, it is best to use custom measures like “EBITDA ex. growth” ignoring expenditures that the company may not need to retain existing customers, annual recurring revenue (“ARR”) based on the latest month or quarter, customer acquisition cost to customer lifetime value (“CAC to LTV”), adjusted earnings, P/E to growth (“PEG ratio”), or many others.  Fundamental financial measures abound, and it is easy to produce near-term estimates for most of them.

For a given fundamental financial measure, which number should one use for the multiple itself?  

It all comes down, of course, to multiples for the comps, the set of companies that one chooses as similar.  What makes each company a “good comparable” to include in a valuation framework? Is it similarity in industry?  Business model? Region? Product mix? Company stage? Growth prospects? Profit margins now or at scale? Target market?  Competitive position? Product features? Long-term contracts? Low churn rate? Patent protection? Management ability to execute?  Access to capital? Scale? Number of competitors? Number of failed competitors? Strategic partners? Regulatory landscape? The comps are the “right” ones only if the characteristics of their outlook are similar to the company that we are valuing.  The real question is: How comparable are the comps, and what adjustments should we make to arrive at the multiple for our company?

Last, will the multiple be the same years in the future when the company needs to raise capital, or when we want to sell our investment to someone else?  Who will be the marginal investor who decides the multiple applied at that future horizon, and what will they base their thinking around at that time? How might the two obvious drivers - comps group and multiples for that comps group - have changed by the next valuation horizon?

How well can we anticipate what multiples will be at an exit years into the future?  

Can we do a better job today of predicting the future multiple by modeling rigorously the actual future cash flows scenarios over the forever horizon so we have a valid snapshot of the company outlook as of any horizon?  Can we predict better how expectations for those future cash flows may change with actual financial results to date, and with certain milestones reached? How can we predict the marginal buyer’s future appetite for comps based on the outlook for the relevant characteristics and the overall sentiment around fear and greed?

Managers and investors generally understand that fear can play a role in lower multiples and greed in higher multiples, and they certainly know that the performance of the business will influence both the chosen fundamental financial measure and the outlook out to the forever horizon, which drives the multiple.

Nonetheless, they feel better avoiding the “false precision” of predicting future cash flows and the “complexity” of considering probability.  With confidence, they predict a near-term financial measure and apply a multiple to compute the entry price paid, and then they live with whatever exit multiple they ultimately receive.  They may run an upside and downside case, or they may briefly consider a sensitivity table of various multiples and various fundamental financial measures, and they discuss various stories that could lead to great, good, or poor outcomes, but curiously they do not focus on the odds of various changes in future cash flows, much less on the odds of various changes in expectations for future cash flows, despite recognition that those expectations influence multiples and valuation.

It sounds hard to estimate the probabilities of changes to actual long-term cash flows.  Forever is a long time, and there are many key drivers that could change a great deal. It sounds even harder to estimate the probabilities of changes to expectations for future cash flows alongside scenarios for milestones reached and for market conditions.  So hard, in fact, that it simply is not practical to go beyond a few multiples or DCF scenarios, or to estimate probabilities even for those few, using spreadsheets, memos, and PowerPoint slides.

However, it is practical with the Bullet Point Network Platform.  We prefer to supplement these traditional valuation methods with a logical set of scenarios that probability-weight how cash flows may actually unfold, what milestones may affect expectations for long-term cash flow at different future horizons, and which events may be conditional on which other events.  We do not just do this one time during the initial underwriting for an investment. We keep that underwriting living and breathing as new evidence becomes available and expectations change.

We believe in Connecting Stories to Statistics to better anticipate changes to exit multiples and to make better investment and strategy decisions. Take a look at the 3 specific steps that follow to see just how you can do so with us. Let’s Connect Stories to Statistics…

Let's Improve the Odds for Future Valuation Multiples by Connecting Stories to Statistics for Future Events

Step 1 pic.PNG

This is the first in a series of three articles exploring how we can improve the odds for future valuation multiples by Connecting Stories to Statistics:

How can statistics for future events help us all improve our decisions?

Our long-term results will be the sum of our results from many decisions whose outcome is impossible to predict with certainty.  Some of these decisions will produce winning outcomes and some will produce losing outcomes, so improving our long-term results relies on increasing the number of winning outcomes and their respective amounts and reducing the number of losing outcomes and their respective amounts.  

Thanks to that simple math, even though we cannot predict each outcome with certainty, we can all improve our long-term results by predicting the odds of each outcome.

Billionaire investor Charlie Munger said it well 25 years ago:

If you don't get this elementary, but mildly unnatural, mathematics of elementary probability into your repertoire, then you go through a long life like a one-legged man in an ass-kicking contest.

How can stories help us all produce better statistics for future events?

Statistics from the past offer a good start, since past odds are observable as frequencies, and many events do repeat, like sunrise and sunset. Old statistical methods like regression analysis can help us extrapolate some of these repeat patterns, and newer methods like deep neural networks can help us extrapolate even more of them. We can all leverage the latest statistical methods with support from leading data science service providers like Two Six Capital.

But billionaire investor Ray Dalio discourages us from extrapolating past data blindly:

The main thrust of machine learning in recent years has gone in the direction of data mining, in which powerful computers ingest massive amounts of data and look for patterns.  While this approach is popular, it’s risky in cases when the future might be different than the past. Investment systems built on machine learning that is not accompanied by deep understanding are dangerous because when some decision rule is widely believed, it becomes widely used, which affects the price.  In other words, the value of a widely known insight disappears over time. Without deep understanding, you won’t know if what happened in the past is genuinely of value and, even if it was, you will not be able to know whether or not its value has disappeared—or worse. It’s common for some decision rules to become so popular that they push the price far enough that it becomes smarter to do the opposite.

On the other hand, in his 2011 New York Times best seller Thinking, Fast and Slow, Economics Nobel Laureate Daniel Kahneman warns us that the fast intuition that Malcolm Gladwell described in his 2007 bestseller Blink is not effective at conceiving statistics for a changing future and that we must develop newer, slower habits of mind to do so.

So are we between a rock and a hard place?

Fortunately, in 2015 Kahneman offered us all some encouragement by writing the forward to the best seller Superforecasting: The Art and Science of Prediction. Its author, University of Pennsylvania Psychology Professor Philip Tetlock, had coined the phrase “The average expert was roughly as accurate as a dart-throwing chimpanzee” in his 2006 book Expert Political Judgment: How Good Is It? How Can We Know? However, in this 2015 book, he encourages us that we can indeed improve our future statistics by supplementing past data with Dalio’s human “deep understanding” through stories:

I realized that as word of my work spread, its apparent meaning was mutating. What my research had shown was that the average expert had done little better than guessing . . . But debunkers go too far when they dismiss all forecasting as a fool’s errand. I believe it is possible to see into the future, at least in some situations and to some extent, and that any intelligent, open-minded, and hardworking person can cultivate the requisite skills.

And Tetlock’s team proved this in its Good Judgment Project, which calculated a Brier score for explicit odds forecasts by over 20,000 people for over 500 future events requested by the CIA’s research arm. It found that the best group of people were 66% more accurate than random guesses through disciplined connection of news stories to statistics. Take that, dart-throwing chimpanzees!

Tetlock offers 10 best practices to join that best group. Notably, Tetlock’s advice there is reinforced by none other than Ray Dalio, whose 2018 best-seller Principles recommends 11 similar best practices for decision-making.

Bullet Point Network serves a client base of elite managers and investors who apply these best practices. They bring to our team companies and investments to work on together confidentially, complementing their deep domain expertise and extensive research by Connecting Stories to Statistics in 3 Steps.

Below, we illustrate Step 1, using the Bullet Point Network Platform to apply each of Dalio’s 11 best practices while Connecting Stories to Statistics for a future event involving a fictitious company that we call Any resemblance to actual companies, living or dead, or actual events is purely coincidental:

1. Recognize that decision-making is a two-step process (first learning and then deciding)

The Platform’s patented architecture includes a Logical Graphical Model that supports the learning step by helping us all connect excerpts from stories more scalably than we can with folders and tags, as well as a Probabilistic Graphical Model that supports the deciding step by helping us all translate those stories into explicit odds that we can weigh.

Logical Graphical Models use Description Logics to infer some qualitative relationships (the dotted lines) from others, helping us draw more connections between excerpts from stories.


Probabilistic Graphical Models use Bayesian Statistics to infer odds of scenarios for some Issues (dotted green and orange) from scenarios for other Issues (yellow) and the conditionality between them (blue and red).


2. Synthesize the situation at hand

When we all identify an important event, like proving out the strength of’s value proposition in its initial market, our Platform’s architecture encourages us all to synthesize data and stories about that event’s potential outcomes from multiple sources and to seek arguments in favor of each of the potential outcomes in order to call their odds more accurately. We can see this below, where the argument from AE, Inc. may carry the most weight in our minds even though it is not the most recent.

vgs first market.PNG

3. Synthesize the situation over time

When we enter new excerpts from stories into the Bullet Point Network Platform, our collaborators who are permissioned for these excerpts can go to our Community Feed and see at a glance how they compare to previous excerpts about the same outcome, so we can all change our odds for that outcome incrementally over time, as Tetlock also advises:

vgs issue news.PNG

4. Navigate levels effectively

Perhaps what we really want to call the odds for is’s peak annual revenue. But to do that, Tetlock advises us to emulate Manhattan Project physicist Enrico Fermi and break peak revenue down successively into what he calls “tractable sub-problems,” such as revenue by market, market share by market, and value proposition by market. As Tetlock says:

The surprise is how often remarkably good probability estimates arise from a remarkably crude series of assumptions and guesstimates.

Our Platform’s graph structure and multiple dimensions make navigating levels more practical for all of us than it is with slides, memos, and spreadsheets:

5. Logic, reason, and common sense are your best tools for synthesizing reality and understanding what to do about it

In the table above, we can see that we do not need some extravagant extrapolation of past statistics in order to handicap many scenarios for peak revenue. Instead, we can do so just by applying logic to what might produce those revenues, such as what might be the number of videogame players in the future and what each might spend on virtual goods, and what percentage via and at what commission rate.

Those are issues for which we can all apply common sense. Especially if we are teenagers or live with some, we can notice when any scenarios portray unrealistic behavior by gamers, even far into the future.

We can also imagine future changes that past statistics alone may not suggest. The late Stanford Investors Professor of Finance Jack McDonald surprised his 1999 MBA investment class on its first day by handing out a newspaper review of the third book in the Harry Potter series, which had been published recently. As his students puzzled over why this was their first handout in an investment class, he challenged them to imagine how big J.K. Rowling could make the Harry Potter franchise. As it turns out, of course, it became so much bigger than the three best-selling children’s books that it was at that time, big as that seemed already. Past statistics did not suggest what Professor McDonald’s imagination did.

6. Make your decisions as expected value calculations

Many business people make decisions based on slides, memos, and spreadsheets that only portray the most-likely case, and perhaps some lip-service to something arbitrarily higher or lower than it. However, the most-likely case is often not representative of the cumulative results of numerous decisions over the long-term. By making it more practical for us to sketch out more scenarios, our Platform helps us see the mean across them so we can make decisions that are more strategic:

7. Prioritize by weighting the value of additional information against the cost of not deciding

One challenge in research is how much to do. We should certainly understand the potential size of relevant market segments as well as the current competitive landscape, but how much time and money should we spend looking into’s value proposition in its first target market, sports videogames published by AE, Inc?

That can depend on how much learning’s value proposition in its first market may influence our odds for its value proposition in broader markets that may take longer to prove out, like all sports videogames and even all videogames. It can also depend on how different categories of’s value proposition, say “Not material” or “Material improvement” or “Radical improvement,” can influence our range for peak market share in each of those markets. By quantifying our research-driven human judgments on these, as Tetlock suggests, we can gain a clear perspective on how important’s value proposition for AE sports videogames may be for its overall peak revenue.

If you would like to explore this perspective on the relationship between initial market value proposition in one chart below and peak revenue in the other chart, you can choose to drag the blue bars or type numbers that add to 100 above the blue bars. If you decide to share the odds that you drag or type, you can click Post Your Own Views below the chart, and then you will be invited to Sign Up to see premium content in our Community Feed:

8. Simplify!

Some might protest that it is too hard to structure in this way everything that we write in a story like an investment memo. That’s right! And that’s the point. Long memos often lull managers and investors into a false sense of security by helping us all feel thorough but failing to connect many of the assertions in our memos clearly to their impact on scenarios for our results.

It is invaluable to force ourselves to apply the 80/20 rule to stories that we write about the future, structuring the minority of the information in them that is likely to influence the majority of our outcomes.

9. Use principles

At one Berkshire Hathaway annual meeting, Warren Buffett was asked whether he would rather hire someone with experience or with the right “latticework of mental models” described by his partner Charlie Munger in his famous 1994 speech, A Lesson on Elementary, Worldly Wisdom As It Relates to Investment Management & Business. Buffett replied that he would like both, of course, but if forced to choose, he would choose the right models over more experience, because we all learn much more from experience when we have the right latticework on which to hang it.

The “elementary, but mildly unnatural, mathematics of elementary probability” is just one of the mental models with which we can learn more from our experience. We must complement it with others. The blog Farnham Street provides a great menu of these.

10. Believability weight your decision making

Those familiar with investing in volatility via financial options may observe that widening our distributions of scenarios is not an adequate solution when our confidence is low, because it may motivate us to overpay for insurance.

Our Platform equips us with another solution, enabling us to assert our level of confidence in each of our judgments and to infer a measure of confidence in other distributions of scenarios implied by those judgments. We can judge our confidence in each judgment based on the quality of the supporting evidence that we have attached directly to it, including whether we have obtained enough information from sources that are believable enough.

Then we can wait to bet only on scenarios for which we have reached reasonable confidence.

11. Convert your principles into algorithms and have the computer make decisions alongside you

In the Bullet Point Network Platform, our team has spent years building scenario models for common phenomena that we all face as managers and investors. We will describe these in subsequent steps:

  • Step 2: Curve Model and Integrated Financial Statements Model

  • Step 3: Bi-Temporal Model of Future Horizons, Capital Markets Rates Model, and Future Multiples Model

  • Bonus: Event Model

As Dalio emphasizes, when we make explicit in a computer model the future scenarios that we conceive, then we learn much more as the future unfolds.

And as Tetlock emphasizes, we can also measure our foresight over time, for example by calculating the percentage of the time the outcomes of events were below the explicit 90th percentile in our forecast from before the event occurred. If we were below our 90th percentile only 50% of the time, then we were too conservative on our upside cases, and if they were below 100% of the time, then we were too aggressive. Either way, misses like these catalyze learning and improving.

The Next Step

Connecting Stories to Statistics for future events is even more useful to our network of managers and investors when we also connect them to statistics for a company’s future cash flows.  We illustrate how in our next article, Step 2: Connect Stories to Statistics for Long-Term Cash Accumulation.

Let's Improve the Odds for Future Valuation Multiples by Connecting Stories to Statistics for Long-Term Cash Accumulation

Step 2 pic.PNG

This is the second in a series of three articles exploring how we can improve the odds for future valuation multiples by Connecting Stories to Statistics:

Invisible Asymptotes

Benchmark Capital co-founder Bill Gurley recommended as “iconic” for “its lucidity, applicability, and therefore overall usefulness,” the concept of “Invisible Asymptotes” described by former Amazon strategic planner Eugene Wei:

It didn't take long for me to see that our visibility out a few months, quarters, and even a year was really accurate (and precise!). What was more of a puzzle, though, was the long-term outlook. Every successful business goes through the famous S-curve, and most companies, and their investors, spend a lot of time looking for that inflection point towards hockey-stick growth. But just as important, and perhaps less well studied, is that unhappy point later in the S-curve, when you hit a shoulder and experience a flattening of growth.

By producing many cases for events as illustrated in our previous article, Step 1: Connect Stories to Statistics for Future Events, we can drive cases for business model characteristics’ long-term asymptotes, such as those below (if you are reading on a smartphone, flipping to landscape mode helps with these interactive charts):

The Bullet Point Network Platform includes a Curve Model with intelligence built in to interpolate realistically from different cases for each characteristic’s long-term asymptote into different cases for its value over all intervening periods, such as years or quarters.

The charts below depict 100 cases for’s future annual revenue in 3 target markets, from first and narrowest to last and broadest, based on the supporting evidence and judgments hinted at in the last article.

Notice that there are many cases where revenue’s asymptote is above zero in the first market segment, reflecting strong odds of a material value proposition there, but there are only about 15 cases where its asymptote is above zero in the last, largest market segment, reflecting only about 15% odds of a value proposition that represents a material improvement or radical improvement for that market..

Please click some curves to explore individual cases:

Long-Term Cash Accumulation

The Platform also includes an Integrated Financial Statements Model with intelligence built in to flex appropriately for different cases for these characteristics in order to produce cases for cash flow in future periods. This makes financial modeling flexible enough to amplify rather than hamstring the strategic thinking prized by managers like Wei and investors like Gurley.

Below, you can see that this is like a typical financial spreadsheet, except that it is not 1 to 3 arbitrary combinations assembled manually but instead 100 cases driven automatically from logical combinations of cases for drivers, with our odds for different combinations suggesting odds for different amounts of long-term cash accumulation. Please drag the dot at the top to explore more of these 100 combinations and their odds:

We can summarize these 100 cases for long-term cash accumulation as curves, which illustrate that cash flow can be more complicated than revenue. Please click some curves to explore individual cases:

In limited situations, cases for a company’s long-term cash accumulation are enough to serve as cases for our long-term return in an investment in the company, such as when . . .

  • Near-term annual cash flow is very high relative to the investment’s price, typically when investors expect that cash flow will crater soon, otherwise known as a “distressed” situation.

  • We have the luxury that our horizon is Warren Buffett’s favorite holding period of “forever.”

On average over the long-term, distressed investors and long-horizon investors have produced more profits than short-term investors, suggesting that it can be very advantageous to have the luxury to operate in these limited situations. Of the hundred wealthiest people in the world as listed by Forbes, only two—Jim Simons and Ray Dalio—are short-term investors, and even those two have made major investments in information systems that they did not sell in the short-term but instead may rely on to help them produce cash flow “forever.”

The Next Step

But hold on.  Distressed high yield situations are few in today’s economic environment, and few of us have the luxury of a horizon as long as Warren Buffet’s “forever.”  Usually we do not operate in these limited situations, so our future returns on an investment are dominated not by its cash flows but by the price at which we can sell it to someone else within our investment horizon.  In those more typical situations, the Bullet Point Network Platform can help us all even more, as we illustrate in our next article, Step 3: Connect Stories to Statistics for Expectations about Long-Term Cash Accumulation at Earlier Horizons.

Let's Improve the Odds for Future Valuation Multiples by Connecting Stories to Statistics for Expectations about Long-Term Cash Accumulation at Earlier Horizons

Step 3 pic.PNG

This is the third in a series of three articles exploring how we can improve the odds for future valuation multiples by Connecting Stories to Statistics:

In our previous article, Step 2: Connect Stories to Statistics for Long-Term Cash Accumulation, we illustrated that it can really pay to call odds on multiple cases for a company’s future long-term cash flows.  But as we noted at the end, most of us need to do more because our limited investment horizons require that we sell our investment in any company long before all its cash flows come to us.  So going from cases for a company’s future cash flows to cases for our own investment returns requires making cases for someone else’s purchase price in the future.

That price is ultimately a human decision that can be influenced by many factors, like fear of missing out (“FOMO”), also known by finance quants as a “momentum factor.”  But whoever is holding the bag at Warren Buffett’s “forever” horizon can simply depend on the company’s long-term future cash flow, so as one steps back to gradually earlier horizons, expectations for that future long-term cash flow maintain a strong magnetism on price, despite the other price influences like FOMO.  In his book Expectations Investing, Columbia Business School Adjunct Professor of Finance Michael Mauboussin points out:

Extensive empirical research demonstrates that the market determines the prices of stocks just as it does any other financial asset. Specifically, the studies show two relationships. First, market prices respond to changes in a company’s cash flow prospects. Second, market prices reflect long-term cash flow prospects . . . companies often need ten years of value-creating cash flows to justify their stock price. For companies with formidable competitive advantages, this period can last as long as thirty years.

Cases for Expectations

Expectations for long-term cash flow can be expressed as stories, as described by New York University Finance Professor Aswath Damodaran in his book Narrative & Numbers: The Value of Stories in Business. He offers an example of our need to think through multiple cases for these stories at our investment horizon:

In my Uber narrative, I viewed Uber as a car service company that would disrupt the existing taxi market (which I estimated to be $100 billion), expanding its growth (by attracting new users) and gaining a significant market share (10 percent). The Gurley Uber narrative was a more expansive one, where he saw Uber’s potential market as much larger (drawing in new users) and its networking effects as much stronger, leading to a higher market share . . . The valuation that I produced for Uber with the Gurley narrative was $28.7 billion, much higher than my estimate of $5.9 billion. Given that the values derived from the narratives were so different, the question, if you were an investor, boiled down to which one had a higher probability of [investors embracing it at your investment horizon].

In fact, MIT Finance Professor Andrew Lo, whose finance research relies heavily on brain research, describes this anticipation of others’ narratives as the lynchpin of anything approaching market efficiency. In his book Adaptive Markets: Financial Evolution at the Speed of Thought, he explains:

The price-discovery process in a well-functioning market requires its participants to engage in a certain degree of cause-and-effect reasoning. “If I do this, then others will do that, in which case I’ll respond by …” This chain of logic presumes that individuals have what psychologists call a theory of mind, the ability to understand another person’s mental state.

We may need to anticipate others’ narratives at multiple future horizons. For example, we may have an expectation that we are confident will make us money within 3 years since the market’s expectation is likely to evolve to ours by then, but it may matter to us what the market’s expectation is in 1 year or 1 quarter because our clients might take their money back if the investment loses over those earlier horizons, robbing us of the chance to win over 3 years even if we were right.

So our added task is to step back from cases for long-term future cash flows to cases for expectations for long-term future cash flows at multiple earlier horizons.  This means 2 time dimensions:  

  1. What eventually happens over future time periods, and

  2. What investors expect for future time periods at each period before they move from future (yellow below) to past (green below).  

2 time dimensions.PNG

Mind blown?  Too hard? When we all reflect, we may note that we do this in our heads, and that is easier than making it explicit in spreadsheets and memos and slide decks, where extra dimensions like this are hard.  But there are advantages to making expectations explicit, which is why managers and investors write spreadsheets and memos and slide decks, and the Bullet Point Network Platform makes it practical to make explicit not only one case for future long-term cash flows but also many cases for future long-term cash flows and even for earlier expectations about that long-term.  That enables us all to amplify what’s in our heads, catching oversights and increasing confidence in our foresight.

We need to think through how cases for each driver from our first article may go from what is priced in today to the cases for what eventually happens.  That involves thinking through serial correlation as well as what Economics Nobel Laureate Daniel Kahneman calls “temporal scope sensitivity.”  The Bullet Point Network Platform’s Bi-Temporal Model of Future Horizons handles these so we all do not have to focus on them but instead can continue to focus on:

  1. What do investors expect now, and

  2. What are the odds for what will eventually happen

The Platform helps us all translate our views on those two points into cases for expectations between those points, which is essential to getting paid within our practical investment horizons.

Remember the previous article’s cases for’s long-term cash accumulation? (If you are reading on a smartphone, flipping to landscape mode helps with the interactive charts below)

The Platform’s second time dimension enables us to consider scenarios not only for what investors will know about the long-term after it happens, like way out in 2050, but also for what they will expect about the long-term before it all happens, like at a target exit horizon in 5 years . . .

. . . and even at more, shorter horizons like 1 year or 2 years, where expectations may influence fundraising potential and dilution:

Cases for the Price Driven by those Expectations

Wait a minute, aren’t we missing something?  How do the expectations at any horizon for long-term cash flows translate into a price at that point?  Is it via a multiple of in year?

There are many ways to fill in those blanks, most of them easy to do in measuring a past price in a past year.  But how do we put odds on cases for how to fill in those blanks at future horizons, especially for growth companies whose outlooks are harder to compare to any real “comparables” than those of more stable, slow-growth companies?  For a price at the 12/31/2023 horizon, why 10x 2024 forward revenue instead of 10x trailing revenue? Or instead of 2x trailing revenue? And do the answers change in cases wherein inflation expectations are 3% annually instead of 2% annually? Or if 10-year Treasury rates are 1% or 5%?

Many articles have been written about how to apply the most universally non-negative multiple, the revenue multiple, to the growthiest of growth companies, tech startups, but most of those articles simply pick today’s revenue multiple for a range of “comparables” and justify it with arbitrary math, claiming without justification that this multiple can be used for more deals at more times.  

A more sincere review of revenue multiples for tech startups reveals giant ranges from 1x to over 600x, and some of the huge differences are between the same company at different times, such as Facebook ranging from under 10x to over 100x.  Another objective article illustrates that even just in a single year, just among private “Unicorns” valued at $1B or more, revenue multiples ranged from 1x to 136x.

And revenue multiples can change radically over time, making it particularly problematic to anticipate future exit multiples mainly based on today’s multiples for comparables. Redpoint Ventures’ Tomasz Tunguz makes cogent observations about valuation multiples in his blog, such as that the entire range of revenue multiples for M&A in 2018 was above the entire range of revenue multiples for M&A in 2012 for the comparables group of Software-as-a-Service companies with over $100M in annual revenue:

Tunguz software multiple ranges.PNG
Tunguz software deals.PNG

Peter Fenton, Bill Gurley’s partner and fellow top-ten member of the 2018 Midas List, has been quoted as saying, “Never turn down a company based on valuation, because it is a mental trap.” When asked to explain this, he detailed that valuation is counterproductive when it is based on arbitrary multiples of the near-term but is actually useful when it is focused instead on the odds of “radical full potential” in the long-term:

There’s probabilistic outcomes and future world states that could occur for every investment . . . You have to be on the field practicing every day, working with companies that have the potential that may never get realized and then intersecting, when all the stars align, the companies where the potential is fully realized . . . What strikes me is that [short-term multiples valuation] thinking so dominates the trade . . . that it sabotages this underlying question of what does it look like in radical full potential, and I think that partners, the venture capitalists that are best suited to work with the company, have to believe in that but also be practical and realistic that there’s a lot of possible future world states and so you have to mitigate that by paying a price that encompasses those potential outcomes.

So how can we all pay a price that encompasses those potential outcomes? Fortunately, the Bullet Point Network Platform can help us all explain consistently when, for example, the odds of long-term full potential might drive investors to pay 10x forward revenue at an investment horizon in 5 years instead of 10x trailing revenue or 2x trailing revenue.

The least arbitrary of multiples, cash flow multiples, are the reciprocal of what real estate investors like to call the “cap rate,” which can be calculated from cost of capital minus expected future growth rate into perpetuity.  The latter term in that calculation is the really vexing one, because it is influenced not just by current annual growth but moreso by the long-term asymptotes emphasized by Wei and Gurley. The key is forecasting expectations for those long-term asymptotes.  We have that key, as we showed in our previous article, Step 2. Anticipate the Odds for Long-Term Cash Accumulation.

So we can predict cases at each future investment horizon for many multiples of many metrics in many years, by:

  • Predicting cases for long-term cash flow expectations at each future investment horizon

  • Putting an “exit multiple” on cash flow only at the long-term asymptotes that end the expectations’ S-Curves.  As Michael Mauboussin explains in Expectations Investing, a company’s cash flow may grow only with inflation beyond those long-term asymptotes, making the cap rate’s “expected future growth rate into perpetuity” term no longer vexing but instead simply the expected inflation rate, for which it is much easier to predict cases.

  • Predicting cases for investors’ opportunity costs, aka “cost of capital.” Predicting a single case for this is easier than predicting a single case for multiple of in year, and predicting the odds of multiple cases for it loosens the reins further.  The Bullet Point Network Platform can make it easy for all of us, since the Platform includes a Capital Markets Rates Model that incorporates . . .

    • The Bullet Point Network co-founders’ decades experiencing the capital markets at senior levels

    • Over a decade of developing patented software architecture and building capital markets models atop it to simulate that experience

    • Ongoing daily work by the Bullet Point Network research team connecting statistical base rates with potential changes suggested by stories in the daily news, providing the continuous and incremental updating that Tetlock advocates

From these ingredients, the Platform’s Future Multiples Model can help us all produce realistic and non-arbitrary cases, at each investment horizon, for the blanks in multiple of in year, and these cases for multiples correspond to the cases for the long-term cash flow expectations, including the asymptotes, that they value.

In the chart below, please click on an EV / revenue curve to see why investors may pay as high as 55x forward annual revenue at a target exit horizon in 5 years, or as low as 2x. These whys help us consider the odds of being above or below any given multiple, which provides invaluable perspective for decision-making. For example, despite a 36x forward revenue multiple today and a high case of a 55x forward revenue multiple in 2023, these whys suggest 75% odds of a multiple below 13x in 2023 and 75% odds of a multiple above 5x in 2023.

These future EV / revenue multiples are not simply a range of current multiples of comparables that may not really be comparable today, much less years into the future. Instead, we have useful probabilities for these multiples, and they are driven logically by cases for capital market conditions, and, most importantly, by expectations for the “radical full potential” to which Fenton refers.

The capability to do this enables us all to focus on our insights about business drivers, which the Platform helps us to translate into realistic scenarios for company cash flows, for investors’ future expectations about those cash flows, and for investors’ future prices for our investments. In the chart below, click the lowest cash in 2020 to see that its case has a high valuation that will make it easy to fund that cumulative cash burn.

Connecting the Stories illustrated in Step 1 to the Statistics illustrated here in Step 3, this analysis suggests a 7-year return median of about 2x and expected value of about 4x. Ray Dalio advises us all to “make decisions as expected value calculations” because expected value represents, much better than the median or most likely case, the cumulative results that we will obtain from numerous strategy and investment decisions over time.

Furthermore, when we express our odds explicitly, then as each horizon moves from future to past, we can assess not just whether any measure, like Net Cash, “beat” or “missed” our outlook but more precisely what our odds were of being above or below it, which can help us to change our outlook more nimbly and also to measure and improve our ability to anticipate odds:

Bullet Point Network serves a client base of elite managers and investors who bring to our team companies and investments to work on together confidentially, Connecting Stories to Statistics in these 3 steps to amplify their long-term returns.

They bring to the BPN team a business plan or investment thesis, plus their research and models that support it.  Our team identifies templates in our Platform that suit the plan or thesis and responds within 5 business days with an Initial Underwriting that Connects Stories to Statistics to amplify the client's work into more scenarios for more confident strategy and investment decisions.  

After the client provides ideas to improve this Initial Underwriting, she works iteratively with the BPN team to produce a Final Underwriting that supports high-confidence decisions.  

Then with the Final Underwriting in our Platform, it continues to live and breathe.  As the environment changes and the client accumulates information, such as around quarterly board meetings, the client and BPN's team can attach research and change judgments regularly to producing Reunderwritings that update the investment's outlook regularly and sincerely, without the enormous work that it would have taken to retrieve a slide deck, memo, and spreadsheet from a folder and rewrite them.

If you might like to try an Initial Underwriting with us, please tell us why below.

Name *

Let's Improve the Odds for Future Valuation Multiples by Connecting Stories to Statistics for Future Events Expectations at Earlier Horizons

Step 4 pic.PNG

This is a bonus addition to the series of three articles exploring how we can improve the odds for future valuation multiples by Connecting Stories to Statistics:

Step 3 illustrated that future investment returns through any investment horizon shorter than Warren Buffett’s “forever” are typically dominated by expectations at that horizon for a company’s long-term cash flows.  Step 2 illustrated that cases for future company cash flow are influenced by the cases for business drivers that Step 1 illustrated.

But we’re still missing something.  Now we are relying on cases for long-term cash flows expectations at earlier horizons, and those are influenced by expectations at earlier horizons for the events that drive those long-term cash flows.

Expectations for events can drive what investors put into their price-setting rationale at an investment horizon.  Typically these expectations are not certain but probabilistic. While investors typically hesitate to write probabilistic expectations explicitly because that is impractical without tools like the Bullet Point Network Platform, some investors are indeed explicit in specific domains where it is more practical because investments are dominated by only a few probabilities.  

Two of these domains are event-driven investing and biotechnology, so let’s start by looking at them and then considering how investors generalize this approach into areas with more uncertainties.

Event-Driven Investing

BPN’s co-founders cut their teeth at Goldman Sachs while its Senior Partner was Robert Rubin, who expanded the firm’s merger arbitrage business into a broader “event-driven investing” style before becoming US Treasury Secretary. Interviewed in Goldman Sachs; The Culture of Success, Rubin said of this investing style:

You had to stick to your discipline and try to reduce everything to plusses and minuses and to probabilities . . . It was a high-risk business, but I’ll tell you, it did teach you to think of life in terms of probabilities instead of absolutes. You couldn’t be in that business and not internalize that probabilistic approach to life.

Event-driven investors try to pay prices explicitly based on their probability of different outcomes to uncertainties like the closing of a merger deal:

Merger arb formula.PNG

Rubin was known for his focus on how future stories might change his reduction of stories to probabilities. His Deputy Treasury Secretary Larry Summers, who later became President of Harvard University, described this focus:

Rubin ends half the meetings with, 'So we don't have to make a decision on this today, do we?' New information will evolve.

We can see this ongoing re-pricing in stock price charts of merger targets as new stories change investors’ reduction to probabilities:

CTB with annotations.PNG


Like event-driven investors, some biotech investors write price targets explicitly based on their probability of the success of future events.  Below, the “Probability-adjustment” of 18% is for the odds of regulatory approval of a company’s drug:

Stock price via EV.PNG

Certainly, to set these probabilities, investors can unpack them into probabilities for earlier events that influence regulatory approval, such as clinical trial phases. There is even abundant historical data on the frequencies of success in each phase conditional on success in the previous phase:

Clinical Development Success Rates.PNG

But each new drug is different from the average drug in its class, so investors seek to apply experience in reducing stories to probabilities as they read and listen to the views of researchers, regulators, insurers, doctors, and patients.

PYBI annotations.PNG

Generalizing this Approach

As illustrated in the first three articles, the Bullet Point Network Platform makes it practical to apply this approach to an investment in any company, estimating how changes in investors’ odds for events may affect the price they will pay for the company’s securities.

But in practice, how can we anticipate ahead of time how investors’ odds for events may evolve over time so we can anticipate how their price may evolve?

Calling Odds on How Odds will Evolve

With the Bi-Temporal Model of Future Horizons that we described in the previous article, we can produce cases for how prospective investors in, for example, may change their expected odds of outcomes and the timing of those outcomes.

To do this realistically in practice, there are nuances to consider, including what billionaire investor George Soros terms, in his 1989 masterwork The Alchemy of Finance: Reading the Mind of the Market, the “reflexivity” between expectations and “actuals” over time, with self-reinforcing and self-defeating cycles between them.  

Atop the Bullet Point Network Platform, we have built an Event Model that considers that nuance and others, making it practical for us all to focus on our views about the eventual odds, applying the best practices detailed in our first article, and then the Event Model will help us explore cases for how expectations may change over time between what is priced in now and what eventually happens.

Below, click on one of the best cases for expected revenue in the lower right to see how it is driven by a case for the change over time in investors’ odds for value proposition strength in each of’s target markets (if you are reading on a smartphone, flipping to landscape mode helps with the interactive charts below):

Together with the Platform’s models for company long-term cash flows and valuation, as illustrated in the previous article, this Event Model enables us all to translate our insights on future events, from product launches to politics, into cases for company cash flow and valuation.

This can enable managers, and the investors who back them, to make better decisions about how much money to raise and to spend based on how and when valuation might change in the future as cash balance changes.  Doing this better repeatedly can make a major improvement in managers’ and investors' career success.

As billionaire investor Peter Thiel writes in his estimable book Zero to One:

You can have agency not just over your own life, but over a small and important part of the world. It begins by rejecting the unjust tyranny of chance. You are not a lottery ticket.

This all starts with organizing stories continually, as illustrated in Step 1, to help us all use our human insight to question and enhance our quantitative scenarios continually. It results in quantification that can help us lean into the home runs and sidestep the train wrecks.

Bullet Point Network serves a client base of elite managers and investors who bring to our team companies and investments to work on together confidentially, Connecting Stories to Statistics in these 3 steps to amplify their long-term returns.

They bring to the BPN team a business plan or investment thesis, plus their research and models that support it.  Our team identifies templates in our Platform that suit the plan or thesis and responds within 5 business days with an Initial Underwriting that Connects Stories to Statistics to amplify the client's work into more scenarios for more confident strategy and investment decisions.  

After the client provides ideas to improve this Initial Underwriting, she works iteratively with the BPN team to produce a Final Underwriting that supports high-confidence decisions.  

Then with the Final Underwriting in our Platform, it continues to live and breathe.  As the environment changes and the client accumulates information, such as around quarterly board meetings, the client and BPN's team can attach research and change judgments regularly to producing Reunderwritings that update the investment's outlook regularly and sincerely, without the enormous work that it would have taken to retrieve a slide deck, memo, and spreadsheet from a folder and rewrite them.

If you might like to try an Initial Underwriting with us, please tell us why below.

Name *